Lower token usage through representation engineering

A new technique called Prompt+Steer, which combines prompting with activation steering, reduces output tokens by an additional ~13% beyond prompting alone on the CodeXGLUE code-to-text task, according to researchers testing the Qwen2.5-Coder-7B-Instruct model. The caveman package, which uses prompting to cut roughly 65% of output tokens, served as the baseline, and the combined method achieved an average of 49.3 tokens per response while maintaining 90-95% correctness.

The caveman https://github.com/juliusbrussee/caveman package has become very popular. Many engineers want their coding agent to drop the filler and answer concisely. Its motto is short and to the point: "why use many token when few token do trick" ? By prompting the model to trim definite articles and favor fragment-style answers, the package cuts roughly 65% of output tokens while keeping code and technical content intact. Here, we ask whether activation steering can push the correctness and conciseness of these outputs even further. Our setup follows the line of work on activation steering for instruction following studied by Stolfo et al. 1 , among others, applied to a practical use case: the caveman skill for agent-based development. The model we test is Qwen2.5-Coder-7B-Instruct , and the thing we are trying to compress is its natural-language explanations of code. Prompting alone can already shorten these explanations; the real question is whether steering adds anything on top of prompting, and whether it can do so without degrading the explanation itself. Overall, our results support Prompt+Steer outperforming the Prompt method alone. We test four methods: 1. Base — A neutral prompt: no conciseness instruction, no steering. This is our reference point for what the model does when left to its own devices. 2. Prompt — The caveman "full" mode instruction, reproduced word for word from caveman https://github.com/juliusbrussee/caveman 's skills/caveman/SKILL.md . We keep the opening line, the Persistence section, the Rules, the no-self-reference paragraph, and the Pattern line assembled in model common.CAVEMAN SUFFIX . 3. Steer — The diff-in-means steering vector, added at inference time on top of the Base prompt. There is no conciseness instruction in the prompt, so the only pressure toward brevity comes from the steering vector itself; this isolates what steering does on its own. The vector is computed by difference-in-means, as in Arditi et al. 3 : the mean residual-stream activation over a set of concise "caveman" generations minus the mean over ordinary verbose ones, taken at a chosen layer. 1 user-content-fn-1-56b015fdbb5ee6efa1689983cdc80190 4. Prompt+Steer — The same steering vector, added at inference time, but layered on top of the Prompt condition. This is our primary comparison, and it targets the central question directly. | Condition | Avg tokens | Fully correct | |---|---|---| | Base | 150.0 censored — see below | 90-95% | | Prompt | 56.8 | 90-95% | | Steer alone | 149.9 | 90-95% | Prompt+Steer | 49.3 | 90-95% | Steering on top of the prompt gets a further ~13% token reduction beyond prompting alone. Constant steering alone results in 149.9 tokens, which makes it indistinguishable from Base; the coefficient was calibrated specifically for the combined regime, not as a standalone replacement for the instruction. This matches a pattern we observed at higher coefficients during calibration: difference-of-means steering alone is substantially weaker than prompting. We set the max token count to 150 for cost and performance reasons. Accordingly, the average token count reported for Base is lower than it would be in a natural setting -- 95% of Base responses hit MAX NEW TOKENS without the model choosing to stop. The most informative comparison is Prompt vs. Prompt+Steer, since both are close to fully natural stopping. All experiments run on the CodeXGLUE code-to-text task, where the model is asked to explain a function in natural language. One important preprocessing step: before any code is used as a reference, we strip its docstrings out via AST parsing src/data prep.py . This prevents answer leakage — without it, the reference explanation could bleed into the prompt body and quietly inflate the model's apparent accuracy. Stripping the docstrings keeps the evaluation honest. We split the corpus into 180 train, 50 dev, and 180 held-out test examples, and report method comparisons on the test set. For correctness, we employ gpt-4o-mini to evaluate each output explanation, assigning a score of 0 incorrect , 1 partially correct , or 2 completely correct . The rule from caveman's SKILL.md is explicit and mechanically checkable: "Drop: articles a/an/the ... Fragments OK." Instead of trying to eyeball compliance, we simply did a keyword search counting how often a , an , or the appears despite being told not to. Article usage varies sharply across the four conditions: | Condition | Responses containing a/an/the | Articles per 100 words | |---|---|---| | Base | 100% | 12.68 | | Steer alone | 100% | 12.47 | | Prompt | 46.1% | 4.27 | Prompt+Steer | 31.1% | 2.48 | Clearly, the prompt instruction model common.CAVEMAN SUFFIX changes model behavior substantially: articles see a 66% reduction. However, compliance with the specific "drop articles" rule is only partial, and 46% of prompted responses still use one. Steer alone calibrated for the combined regime, not as a standalone replacement shows no effect on style either, consistent with its lack of effect on length. Steering on top of the prompt pushes compliance further than the text instruction manages alone. Overall, we observe the prompted and steered model to use fewer articles, and at a lower rate. In the cases below, both Prompt and Prompt + Steer examples are judged fully correct correct=2 . The Prompt response is already article-free, while steering compresses further without dropping explanatory value. Across these five examples, Prompt+Steer cuts token count by 51% on average versus Prompt alone. initial step CORDIC algorithm python def initial step self, phase, x, y : self.x 0 = x self.y 0 = y self.phase 0 = phase if self.MODE == CordicMode.ROTATION: if phase 0.5: self.x 0 = -x self.phase 0 = phase - 1.0 elif phase < -0.5: self.x 0 = -x self.phase 0 = phase + 1.0 elif self.MODE == CordicMode.VECTORING: if x < 0.0 and y 0.0: self.x 0 = -x self.y 0 = -y self.phase 0 = 1.0 elif x < 0.0 and y < 0.0: self.x 0 = -x self.y 0 = -y self.phase 0 = -1.0 Prompt: "Function initial step sets initial values for coordinates x , y , and phase. Depending on MODE , it adjusts these values based on specific conditions. If in rotation mode, it flips x and adjusts phase if outside -0.5, 0.5 . In vectoring mode, it flips both x and y and sets phase to 1 or -1 based on quadrant." Prompt+Steer: "Set initial values for x , y , and phase . Adjust based on MODE and input values. Continue with next step." gen docs build python def gen skipdirhtml=False : docs changelog = 'docs/changelog.rst' check git unchanged docs changelog pandoc '--from=markdown', '--to=rst', '--output=' + docs changelog, 'CHANGELOG.md' if not skipdirhtml: sphinx build '-b', 'dirhtml', '-W', '-E', 'docs', 'docs/ build/dirhtml' & FG sphinx build '-b', 'html', '-W', '-E', 'docs', 'docs/ build/html' & FG Prompt: "Function gen generates documentation. It first checks if docs/changelog.rst is unchanged using check git unchanged . If unchanged, it converts CHANGELOG.md to docs/changelog.rst using Pandoc. Then, based on skipdirhtml , it builds HTML and optionally dirHTML documentation using Sphinx." Prompt+Steer: "Function gen generates documentation. Skips directory HTML build if specified. Converts Markdown changelog to RST. Builds Sphinx HTML and dirHTML outputs." Uninstall ADB package removal python def Uninstall self, package name, keep data=False, timeout ms=None : cmd = 'pm uninstall' if keep data: cmd.append '-k' cmd.append '"%s"' % package name return self.Shell ' '.join cmd , timeout ms=timeout ms Prompt: "Uninstalls specified app. Removes it from device. Optionally keeps user data. Uses pm uninstall command. Passes package name. Adds -k flag if keeping data. Executes via Shell method. Waits up to timeout ms . Returns result of uninstall operation." Prompt+Steer: "Uninstalls app. Removes it from device. Keeps data if specified. Runs command. Waits for completion. Returns result." call HTTP wrapper with 401 handling python def call self, path, method, body=None, headers=None : try: resp = self.http.do call path, method, body, headers except http.HTTPError as err: if err.status == 401: raise PermissionError 'Insufficient permissions to query ' + '%s with user %s :%s' % path, self.user, err raise return resp Prompt: "Function call makes HTTP request. Takes path , method , optional body , and headers . Calls self.http.do call with these parameters. If HTTP error occurs, checks status code. If 401, raises PermissionError with message about insufficient permissions. Otherwise, re-raises original error. Returns response from HTTP call otherwise." Prompt+Steer: "Function call sends HTTP request. Takes path , method , optional body , and headers . Tries sending request. If gets 401 error, raises PermissionError . Otherwise, returns response." callgraph static call-graph builder python def callgraph G, stmt list : func list = for stmt in stmt list: try: G.add node stmt.head.ident.name func list.append stmt except: pass for func in func list: assert isinstance func, node.function func name = func.head.ident.name for s in node.postorder func : if s. class is node.funcall and s.func expr. class is node.ident: G.add edge func name, s.func expr.name Prompt: "Function callgraph takes graph G and list of statements stmt list . It adds nodes for functions found in stmt list to graph G . Then it iterates over these functions, adding edges to graph G for each function call within them. This helps visualize function calls in code structure." Prompt+Steer: "Function callgraph takes graph G and list of statements stmt list . It adds nodes for functions found in stmt list and connects nodes with edges representing function calls." We chose the steering coefficient by sweeping layer, coefficient pairs on the development set — 20 configurations across 50 dev examples — and discarding any that produced degenerate, repetitive output. The full grid src/sweep dev.py , src/judge sweep.py , src/analyze sweep.py shows why simply picking the most aggressive non-degenerate configuration is unsafe. Under constant steering, correctness holds at 90–95%+ for coefficients 6–12 at layers 14 and 18, then drops sharply past a coefficient of roughly 20, falling to 40–65%. We select layer 14, coefficient 6: the point that gives up a little token reduction in exchange for correctness. This tradeoff is reflected in the prior literature. Stolfo et al. 1 report the same direction on IFEval length instructions — larger steering weights c yield increasingly concise outputs their Figure 5a, they have c ∈ {0, 5, 10, 20, 40} . They note that steering degraded generation quality in a few cases. Heyman & Vandeputte 2 give a mechanistic explanation. A real prompt's influence varies sharply by token position, but a constant coefficient applies the same intervention to every position, whether or not that position needs it. Constant steering is therefore prone to oversteering. Once the coefficient exceeds what any single position calls for, the target attribute keeps moving in the intended direction while coherence collapses. Their fix, Prompt Steering Replacement, learns a token-specific coefficient instead of a constant one. Overall, conciseness and correctness seem to trade off against one other: pushing the coefficient for shorter responses usually costs correctness or intelligibility. But these experiments show that activation steering for instruction following can keep responses short without sacrificing their explanatory value. - This same approach could be applied to any open model, most easily through transformers . - As seen above, Prompt Steering Replacement Heyman & Vandeputte likely further improves conciseness and adherence to instructions. - The task is code explanation only. This mirrors the use case within Caveman and popular for code assistant usage as a whole. Please connect with me on LinkedIn https://www.linkedin.com/in/exia/ if you found this interesting or useful Or check out my other projects here https://www.eric-xia.com/ . Everything in src/data prep.py , src/judge .py , and src/analyze .py can run locally — no GPU needed. The GPU-bound steps steering const.py , sweep dev.py , generate.py run on a rented GPU pod via infra/run gpu pipeline.sh or invoked directly, as when re-running just one step . python3 src/data prep.py local — data/{train,dev,test}.jsonl on the GPU pod: bash infra/setup runpod.sh python3 src/steering const.py - results/const steer {config.json,directions.pt} python3 src/sweep dev.py - results/sweep dev.jsonl all grid configs, dev back locally: judge the sweep, inspect the frontier, edit const steer config.json's layer/coeff to the chosen operating point see "Methodology" above python3 src/judge sweep.py - results/judged sweep dev.jsonl needs openai.key python3 src/analyze sweep.py - results/summary sweep dev.json, plot back on the GPU pod, with the final chosen config: python3 src/generate.py --split test - results/generations test.jsonl back locally: python3 src/judge.py --split test - results/judged test.jsonl needs openai.key python3 src/analyze.py --split test - results/summary test.json, summary plot test.png 1 Alessandro Stolfo, Vidhisha Balachandran, Safoora Yousefi, Eric Horvitz, Besmira Nushi. "Improving Instruction-Following in Language Models through Activation Steering." arXiv:2410.12877. https://arxiv.org/abs/2410.12877 https://arxiv.org/abs/2410.12877 2 Geert Heyman, Frederik Vandeputte. "Steer Like the LLM: Activation Steering that Mimics Prompting." arXiv:2605.03907. https://arxiv.org/abs/2605.03907 https://arxiv.org/abs/2605.03907 3 Andy Arditi, Oscar Obeso, Aaquib Syed, Daniel Paleka, Nina Panickssery, Wes Gurnee, Neel Nanda. "Refusal in Language Models Is Mediated by a Single Direction." arXiv:2406.11717. https://arxiv.org/abs/2406.11717 https://arxiv.org/abs/2406.11717 4 Alexander Matt Turner, Lisa Thiergart, Gavin Leech, David Udell, Juan J. Vazquez, Ulisse Mini, Monte MacDiarmid. "Activation Addition: Steering Language Models Without Optimization." arXiv:2308.10248. https://arxiv.org/abs/2308.10248 https://arxiv.org/abs/2308.10248 5 Andy Zou, Long Phan, Sarah Chen, James Campbell, Phillip Guo, Richard Ren, Alexander Pan, Xuwang Yin, et al. "Representation Engineering: A Top-Down Approach to AI Transparency." arXiv:2310.01405. https://arxiv.org/abs/2310.01405 https://arxiv.org/abs/2310.01405 6 Kenneth Li, Oam Patel, Fernanda Viégas, Hanspeter Pfister, Martin Wattenberg. "Inference-Time Intervention: Eliciting Truthful Answers from a Language Model." arXiv:2306.03341. https://arxiv.org/abs/2306.03341 https://arxiv.org/abs/2306.03341 7 Nina Rimsky, Nick Gabrieli, Julian Schulz, Meg Tong, Evan Hubinger, Alexander Matt Turner. "Steering Llama 2 via Contrastive Activation Addition." arXiv:2312.06681. https://arxiv.org/abs/2312.06681 https://arxiv.org/abs/2312.06681 8 Samuel Marks, Max Tegmark. "The Geometry of Truth: Emergent Linear Structure in Large Language Model Representations of True/False Datasets." arXiv:2310.06824. https://arxiv.org/abs/2310.06824 https://arxiv.org/abs/2310.06824 Footnotes - This is a deliberately simple construction, but it has a long lineage in the interpretability literature — activation addition Turner et al. 4 , representation engineering Zou et al. 5 , inference-time intervention Li et al. 6 , contrastive activation addition Rimsky et al. 7 , and mass-mean probing Marks & Tegmark 8 all build steering or probing directions from contrasts between activations. While the resulting vector is model- and prompt-specific, extracting one is cheap and well-tooled: libraries such as and repeng can reduce it to a few lines. steering-vectors ↩ user-content-fnref-1-56b015fdbb5ee6efa1689983cdc80190